Muscle-Synergy-Based Planning and Neural-Adaptive Control for a Prosthetic Arm

Upper limb loss has significant effects on the individual's quality of life. Artificial prosthetic limbs as an alternative to the lost limb are designed to allow amputees to regain motor function. Motion classification via extracted surface electromyogram (sEMG) signals is widely utilized to realize a friendly human-robot interface in the control of the prosthesis. However, limited classification of discrete motion patterns from sEMG prevents intuitive motor control. Thus, instead of using discrete patterns, decoding the human intention continuously from sEMG would significantly benefit the prosthesis control. In this article, we propose a muscle-synergy-based intention decoding and motion planning that can model a broad set of complex upper limb movements as a combination of motor primitives. A novel muscle activation-to-force mapping model is developed to detect muscular effort of the healthy side to drive the affected side. A neural-network-approximation-based controller is developed for the bionic neuroprosthetic arm to execute the movement. Operational experiments with prosthetic movement control were performed on four healthy participants and an upper limb amputee participant. Results of controlling prosthetic arm (0.94 ± 0.02 of R2 in the horizontal reaching tasks and 0.95 ± 0.01 in the vertical reaching tasks for the healthy subjects, 0.95 and 0.97 for the amputee) demonstrate that our control method could successfully capture human movement intention and effectively control the movement of prosthesis. Impact Statement-Powered robotic prostheses have enabled individuals with limb amputations to regain motor function via motion intention decoding from extracting surface electromyogram signals. Most research works have provided human-robot interaction through motion classification obtained from pattern recognition using surface electromyogram signals and simple movement control. However, decoding the human intent continuously from surface electromyogram is considered to be an ideal method to realize intuitive and natural prosthetic control for the user. Muscle synergies consider using the underlying coordination principles to extract human intention from surface electromyogram as a biologically plausible approach to simplify the muscle redundancy problem. In this article, a muscle-synergy-based control framework is proposed, which combines intention decoding with motion control. © 2021 IEEE.